Intelligent Tutoring Systems (ITS) can be very effective in improving student learning [1, 2]. They can also serve as useful platforms for controlled experiments in learning science: researchers can alter a tutor to include a new instructional intervention and then administer altered and unaltered versions of the tutor to their subjects. However, the difficulty of creating tutoring systems, which traditionally have required cognitive modeling and artificial intelligence programming skills, has both hindered their acceptance in mainstream education and made them costly vehicles for researchers. [3]. The main goal of the Cognitive Tutor Authoring Tools (CTAT) project is to provide a suite of authoring tools that make tutor development more affordable by leveraging human-computer interaction (HCI), machine learning, and data mining techniques. Previous work on CTAT added the capability for nonprogrammers to create example-tracing tutors via a programming-by-demonstration technique that required no coding [4]. While example-tracing tutors provide a student experience similar to that of the more general cognitive tutors, they also require that an author demonstrate and fully annotate (i.e., provide error messages and hints) each individual problem to be presented.

Prior CTAT tutorials emphasized programming by demonstration for single tutors and tools for creating cognitive models based production-rule systems [5]. The tutorial proposed here will include example-tracing tutors but also present enhancements suggested by use of the tools by projects in the Pittsburgh Science of Learning Center (http://www.learnlab.org), a large, NSF-funded center in the United States.

One set of new capabilities is intended to permit so called "mass production" of example-tracing tutors. Previously, correct student inputs and expected errors were constants in the solution graph that authors recorded for each problem instance. The mass-production enhancement permits variables in the example-tracer’s solution graph, so that a single graph becomes a template for any number of problems. Now, individual problem values can be entered in the widely-used off-the-shelf Excel spreadsheet program. Furthermore, authors can use Excel’s own programming and external querying capabilities to fill in values for specific problems.

The tutorial is meant especially for the following categories of participants:

1. Developers interested in integrating "learning by doing" components in online educational systems and looking for tools to make that easier.
2. Researchers who want to quickly and cheaply generate, test and deploy different pedagogical interventions in tutoring systems in order to explore hypotheses about learning or instruction.
3. Educators (e.g., college-level professors) with minimal technical background interested in developing on-line exercises for their courses.

Objectives:

By direct, hands-on experience, the tutorial seeks to enable participants to evaluate whether the CTAT tools would be useful for their own development, research or teaching activities. Participants also will learn some of the general capabilities of intelligent tutoring systems and their usefulness to education and research.

A second objective is for participants to reflect and comment on the tools’ usefulness and usability, especially with respect to their new capabilities. We seek ideas that might render CTAT more intuitive, less labor-intensive and more broadly applicable.